I like to fine-tune aileron differential for aerobatics by the same method Baron Johnson described. I get the model really moving directly away from me, then give it full aileron. As it rolls, I move the elevator control slightly up and down, and observe whether the roll looks more axial with a little elevator input. If the roll looks best with a little down elevator, it means that the airplane is barreling inside, so it needs a little positive differential (more up than down). I add a little positive differential, and repeat the same test. Sometimes a little negative differential is needed.